The role of GDNF in patterning the excretory system

Transcription

1 Developmental Biology 283 (2005) The role of GDNF in patterning the excretory system Reena Shakya a, Eek-hoon Jho a,1, Pille Kotka a,2, Zaiqi Wu a, Nikolai Kholodilov b, Robert Burke b,c, Vivette D Agati c, Frank Costantini a, * a Department of Genetics and Development, Columbia University Medical Center, 630 W. 168th Street, New York, NY 10032, USA b Department of Neurology, Columbia University Medical Center, 630 W. 168th Street, New York, NY 10032, USA c Department of Pathology, Columbia University Medical Center, 630 W. 168th Street, New York, NY 10032, USA Received for publication 24 January 2005, revised 28 March 2005, accepted 6 April 2005 Available online 10 May 2005 Abstract Mesenchymal epithelial interactions are an important source of information for pattern formation during organogenesis. In the developing excretory system, one of the secreted mesenchymal factors thought to play a critical role in patterning the growth and branching of the epithelial ureteric bud is GDNF. We have tested the requirement for GDNF as a paracrine chemoattractive factor by altering its site of expression during excretory system development. Normally, GDNF is secreted by the metanephric mesenchyme and acts via receptors on the Wolffian duct and ureteric bud epithelium. Misexpression of GDNF in the Wolffian duct and ureteric buds resulted in formation of multiple, ectopic buds, which branched independently of the metanephric mesenchyme. This confirmed the ability of GDNF to induce ureter outgrowth and epithelial branching in vivo. However, in mutant mice lacking endogenous GDNF, kidney development was rescued to a substantial degree by GDNF supplied only by the Wolffian duct and ureteric bud. These results indicate that mesenchymal GDNF is not required as a chemoattractive factor to pattern the growth of the ureteric bud within the developing kidney, and that any positional information provided by the mesenchymal expression of GDNF may provide for renal branching morphogenesis is redundant with other signals. D 2005 Elsevier Inc. All rights reserved. Keywords: Organogenesis; Kidney development; GDNF; Ret receptor tyrosine kinase; Ret; Hoxb7; Metanephric kidney; Wolffian duct; Ureteric bud; Branching morphogenesis; Metanephric mesenchyme Introduction Mesenchymal epithelial interactions are thought to play a critical role in the growth and pattern formation of organs that develop by branching morphogenesis, such as the kidney and lung. During kidney development, the epithelial ureteric bud (UB) emerges from the Wolffian duct, in response to inductive signals from the metanephric mesenchyme, a specialized region of the nephrogenic cord (Saxen, * Corresponding author. Fax: address: fdc3@columbia.edu (F. Costantini). 1 Present address: Department of Life Science, University of Seoul, 90 Jeonnong-Dong, Dongdaemun-Gu, Seoul, Korea. 2 Present address: National Institute of Chemical Physics and Biophysics and Tallinn University of Technology, Department of Gene Technology, Tallinn, Estonia. 1987). As the UB invades the metanephric mesenchyme, further inductive signals from the mesenchyme and renal stromal cells (Carroll and McMahon, 2003; Shah et al., 2004; Vainio and Lin, 2002) induce the UB to undergo a complex and stereotypic process of branching and elongation, eventually giving rise to the epithelium of the ureter, renal pelvis, calyces, and collecting ducts (Al-Awqati and Goldberg, 1998; Oliver, 1968). The ureteric bud, in turn, secretes factors that induce the metanephric mesenchymal cells to condense, convert to epithelia and eventually form the various segments of the nephron (Carroll and McMahon, 2003; Saxen, 1987). One of the main signaling pathways that promotes ureteric bud branching morphogenesis involves the secreted factor GDNF, which signals through the Ret receptor tyrosine kinase and the co-receptor GFRa-1 (Sariola and /$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi: /j.ydbio

2 R. Shakya et al. / Developmental Biology 283 (2005) Saarma, 2003). Prior to outgrowth of the UB from the Wolffian duct, Ret and GFRa-1 are expressed throughout the Wolffian duct, while GDNF is expressed broadly in the nephrogenic cord surrounding the posterior duct (Hellmich et al., 1996; Pachnis et al., 1993; Suvanto et al., 1996). As the UB evaginates from the caudal end of the Wolffian duct, the expression of GDNF becomes limited to and upregulated in the adjacent metanephric blastema (Grieshammer et al., 2004). Concomitantly, Ret and GFRa-1 are downregulated in the Wolffian duct and become restricted to the newly formed ureteric bud. As UB branching progress, Ret and GFRa-1 soon become restricted to the tips of UB branches, and GDNF to the surrounding mesenchymal cells at the periphery of the kidney (Pachnis et al., 1993; Sainio et al., 1997). The importance of this pathway for renal development was established by the finding that null mutations in gdnf, ret, or gfra-1 result in renal agenesis or severe dysgenesis, due to the failure of the UB to emerge from the Wolffian duct or to grow and branch normally (Cacalano et al., 1998; Enomoto et al., 1998; Moore et al., 1996; Pichel et al., 1996; Sanchez et al., 1996; Schuchardt et al., 1994, 1996). However, the specific cellular and developmental processes induced by GDNF remain unclear. While there were conflicting reports as to whether GDNF induces UB cell proliferation (Michael and Davies, 2004; Pepicelli et al., 1997; Sainio et al., 1997; Vega et al., 1996), a recent analysis of the behavior of ret / cells in mosaic kidneys supports the conclusion that Ret signaling is particularly important for the proliferation of cells at the growing UB tips (Shakya et al., 2005). GDNF may also promote the survival of UB cells (Towers et al., 1998), as it does for some neuronal populations (Sariola and Saarma, 2003). A third potential role, for which there is considerable evidence, albeit indirect, involves specifying the pattern of branching morphogenesis. A GDNF-soaked bead can induce the outgrowth of ectopic buds from the Wolffian duct in organ culture (Brophy et al., 2001; Pichel et al., 1996; Sainio et al., 1997), while mutations that alter the distribution of GDNF along the Wolffian duct can cause ectopic UBs to form in vivo (Grieshammer et al., 2004; Kume et al., 2000). Addition of GDNF to kidney cultures can cause increased or abnormal UB branching (Pepicelli et al., 1997; Pichel et al., 1996; Sainio et al., 1997; Towers et al., 1998; Vega et al., 1996), while expression of activated forms of Ret in the Wolffian duct and ureteric bud also induces ectopic UBs and abnormal patterns of UB growth and branching (de Graaff et al., 2001; Srinivas et al., 1999b). These studies have led to the idea that GDNF/Ret signaling may specifically induce branching of the UB epithelium (i.e., that it is a ramogen ), and that it provides a chemoattractive cue, whose concentration gradient is sensed by the UB tips and thus directs their growth. The latter hypothesis was supported by studies in which Ret-expressing cells were shown to migrate towards a source of GDNF (Tang et al., 1998), as well as by observations that GDNF may serve as a chemoattractant for migrating enteric neural crest cells (Natarajan et al., 2002; Young et al., 2001). Thus, a model has emerged in which the localized expression of GDNF is important for normal patterning of the excretory system, first in determining the site of initial outgrowth of the UB, and later in determining the branching pattern within the developing kidney (Lechner and Dressler, 1997; Sariola and Saarma, 2003; Sariola and Sainio, 1997). Another potential role of GDNF involves the proximal distal patterning of the ureteric bud. GDNF is expressed only at the periphery of the growing kidney, and several genes expressed specifically in the peripheral UB tips, including Ret and Wnt11, are upregulated by GDNF (Pepicelli et al., 1997). Thus, it is possible that the pattern of expression of these (and perhaps other) tip-specific genes is a direct consequence of the localized expression of GDNF. To test these hypotheses in an in vivo situation, we generated transgenic mice in which GDNF is ectopically expressed in cells throughout the Wolffian duct and ureteric bud epithelium. The effects of GDNF misexpression were examined first on a wild type background, in which endogenous GDNF is also expressed by the mesenchyme, and then on a gdnf knockout background, in which endogenous GDNF is absent, and replaced by the ectopically-expressed GDNF. We reasoned that if GDNF were required only for UB cell survival and proliferation, such alterations in its spatial distribution might have little impact on the pattern of UB outgrowth and branching; however, if GDNF served as a critical chemoattractive factor, they should have profound effects on UB branching morphogenesis. Our results support the conclusion that GDNF stimulates bud formation by the Wolffian duct epithelium and that the localized expression of GDNF in the posterior nephrogenic cord is important for the outgrowth of a single, correctly placed ureteric bud. They also show that growth and branching of the collecting duct system within the kidney can be perturbed to some extent by the misexpression of GDNF in the ureteric bud. Surprisingly, many aspects of kidney development and UB morphogenesis were normal in mice in which GDNF was expressed only in the Wolffian duct and UB epithelial cells, but not in the mesenchyme. This observation argues that, if the mesenchymal expression of GDNF does provide positional cues for ureteric bud growth and patterning, this information must be redundant with other signals that can independently pattern the ureteric bud. Materials and methods Generation of Hoxb7/rtTA transgenic mice The Hoxb7 promoter ( 1316 bp to +81 bp of the Hoxb7 gene) (Kress et al., 1990) was excised from

3 72 R. Shakya et al. / Developmental Biology 283 (2005) pgem-blue with KpnI and AvaI, the ends blunted, and subcloned into the EcoRV site of pbluescriptii KS+. The Hoxb7 promoter in pbluescript was then excised with XhoI and EcoRI and the EcoRI end blunted. Plasmid puhdrtta2s-m2 (Urlinger et al., 2000) was digested with XhoI and SacI to remove the CMV promoter, leaving the rtta2 S -M2 gene followed by SV40 polya sequences. The SacI site was blunted, and the Hoxb7 promoter fragment was then cloned into the XhoI SacI fragment of the puhdrtta2s-m2 vector. The entire 2.7 kb Hoxb7/rtTA2S-M2/SV40-pA transgene (abbreviated Hoxb7/rtTA ) was excised from the vector with XhoI and HindIII for pronuclear injection into B6CBAF1/J mouse eggs (Hogan et al., 1994). Three transgenic founder animals (RS-HTA2, RS-HTA3 and RS-HTA8) were identified by Southern blot using EcoRI, which cuts once within the transgene, using a probe generated from the rtta2 S -M2 cdna. Offspring of the three founders were tested for expression of rtta2 S -M2 in kidney by RT-PCR (data not shown) and by crossing to teto/h-gal reporter mice, inducing with Dox and staining for h-gal (e.g., Figs. 2C, D, 3A and data not shown). All experiments in this paper were conducted using animals from line RS-HTA2, on a mixed genetic background. Inheritance of the transgene was monitored by PCR using primers rtta2s-a (5V CATGGCAAGACTTTCTGCGG) and rtta2s-b (5VTTGQTCTCAGAAGTGGGGGCA), which amplified a band of 296 bp from the rtta2 S - M2 sequence under the following conditions: 94-C, 5 min; 30 cycles of (94-C, 30 s; 58-C, 30 s; 72-C, 50 s); 72-C, 7 min. Generation of Hoxb7/GDNF transgenic mice A PmeI restriction site was introduced 5Vof the Hoxb7 promoter in pbluescriptii (KS+) (see above) while a multiple cloning site (MCS) (5V EcoRI HindIII PacI NheI KpnI XhoI AseI NotI BamHI PstI PmeI 3V) was introduced downstream of the Hoxb7 fragment. A BamHI PstI fragment including an intron and polyadenylation sequences of the human h-globin gene was subcloned into the MCS to generate plasmid Hoxb7- MCS-hh-globin-pA. The mouse GDNF cdna ( 636 bp) was excised from plasmid GDNFXVL with HindIII and partial BamHI digestions, HindIII end blunted, and subcloned into EcoRI (blunted) and BamHI sites, upstream of an IRES2-EGFP sequence in pbluescriptii (KS+). The GDNF-IRES2-EGFP fragment was excised with NheI and NotI, the NotI end blunted, and subcloned into the NheI and BamHI (blunted) sites in the MCS of Hoxb7-MCS-hhglobin-pA. The final construct, RS#29, was verified by restriction mapping and partial sequencing. The 5 kb insert was excised with PmeI, gel-purified, and eluted in injection buffer (Hogan et al., 1994). Transgenic fetuses were identified by Southern blot using EcoRI and a mouse GDNF cdna probe (data not shown). Fig. 1. Excretory system defects in Hoxb7/GDNF transgenic mice. (A) Schematic diagram of transgene construct. (B) and (D) Sagittal sections of two transgenic kidneys at E18.5. Asterisks denote cystic UB tips in the cortex. (C) and (E) Enlargements of boxed areas in B and D, showing branching of cyst epithelium (arrows). Inset in C, dolichos biflorus lectin (brown) staining confirms the UB origin of cystic epithelia; methyl green counterstain. Inset in E shows hyperplasia of the cystic epithelium, which is 5 6 cell diameters in thickness compared to the simple epithelium of normal collecting ducts. (F) The male reproductive tract is abnormally dilated (e, epididymis) and branched ectopic UBs are found nearby (arrowheads). t, testis; b, bladder. (G) Enlargement of the boxed area in (F) showing extensive branching of ectopic UBs. H and E staining. Scale bars, 200 Am.

4 R. Shakya et al. / Developmental Biology 283 (2005) TA2 S -M2 (a gift of Dr. Wolfgang Hillen) was replaced with the 5.7-kb Axin2 fragment, using the enzymes described above for Hoxb7/rtTA. The Axin2-rtTA plasmid was digested with AflII and HindIII and the 7-kb insert purified for microinjection. Genotyping was performed with the same PCR primers as for Hoxb7/ rtta. BiTetO/lacZ/GDNF and teto/lacz transgenic mice and GDNF mutant mice Strain BiTetO/lacZ/GDNF has been described (Kholodilov et al., 2004). The teto/lacz strain B6;SJL-Tg(tetoplacZ)2Mam/J (Furth et al., 1994) was obtained from The Jackson Laboratory. The gdnf lacz knock-in null allele (Sanchez et al., 1996) was obtained from M. Barbacid (Bristol-Meyers Squib). Transgene induction with Doxycycline Fig. 2. Transgene constructs for inducible expression of GDNF and h-gal in the Wolffian duct and ureteric bud. A C, transgene constructs. A, Hoxb7/ rtta2 S -M2 (abbreviated Hoxb7/rtTA) or Axin2/rtTA2 S -M2 (abbreviated Axin2/rtTA). B, BiTetO/lacZ/GDNF (Kholodilov et al., 2004). C, teto/lacz (Furth et al., 1994). D, expression of control teto/lacz transgene, activated by Hoxb7/rtTA, in the renal collecting ducts of a 40-day-old bi-transgenic mouse maintained on Dox. c, cortex; m, medulla; p, papilla. Generation of Axin2-rtTA transgenic mice The 5.7-kb insert of plasmid Axin2P-Luc (Jho et al., 2002), containing the Axin2 promoter was excised with MluI and HindIII, and the CMV promoter in puhdrt- Pregnant females were fed 2 mg/ml Doxycycline (Sigma D-9891), 5% sugar (w/v) in the water. Dox was administrated continuously from day 8.5 of gestation in the experiments of Figs. 3 6, from day 9.5 or 10.5 in the rescue experiments (Figs. 7 and 8). Full induction is achieved 1 day after the start of Dox (Kistner et al., 1996). The rationale for delaying the induction in the rescue experiments was to wait for a UB to form naturally ( E11), as occurs in some gdnf / embryos, and then support its further growth by the induction of the transgene-encoded GDNF. Whether this delay contributed to the success of the rescue is not known. Fig. 3. Ectopic ureteric buds emerging from the Wolffian duct of Hoxb7/rtTABiTetO/lacZ/GDNF bi-transgenic embryos at E11.5. A, control Hoxb7/ rttateto/lacz bi-transgenic embryo with a single T-stage UB (arrow) that emerged from the posterior end of the Wolffian duct. B, a weakly affected Hoxb7/ rttabiteto/lacz/gdnf bi-transgenic embryo with a secondary T-stage UB (red arrow) anterior to the normal UB (black arrow), and a few small buds along the more anterior Wolffian duct (yellow arrows). C, a strongly affected Hoxb7/rtTABiTetO/lacZ/GDNF bi-transgenic embryo with multiple UBs all along the length of the Wolffian duct. Ectopic buds in more anterior regions of the Wolffian duct had not yet branched at E11.5. Scale bar, 0.5 mm.

5 74 R. Shakya et al. / Developmental Biology 283 (2005) b-gal staining of whole mount tissues and cryosections Tissues were fixed in fresh, ice-cold 2% PFA (paraformaldehye) and 0.2% glutaraldehyde in 1 PIPES buffer (0.1 M Pipes, 2 mm MgCl 2, 2 mm EGTA, ph 6.9) 4-C for 1 h, washed in PBS, and postfixed in 0.2% glutaraldehyde in 1 PIPES buffer for 1 h at room temp. Alternatively, tissues were fixed in 4% PFA in PBS for 1 3 h at 4-C. After

6 R. Shakya et al. / Developmental Biology 283 (2005) several washes in PBS, for sectioning tissues were permeated with 15% and 30% sucrose in 0.1 M phosphate buffer, embedded in OCT (Tissue-Tek \ ) and frozen in dry ice Am sections were fixed in 2 mm MgCl 2 and 0.2% glutaraldehyde in PBS for 5 min before staining with X-gal. Sections and whole tissues were stained with fresh X- gal (2 mm MgCl 2, 2.12 mg/ml potassium ferrocyanide, 1.64 mg/ml potassium ferricyanide, 1 mg/ml X-gal in PBS). Tissues were stored in 10% Formalin buffer, while cryosections were mounted with non-aqueous mounting medium. In situ hybridization In situ hybridization was performed with digoxigeninlabeled riboprobes as described (Mendelsohn et al., 1999) and on whole-mount tissues as described (Wilkinson, 1992). Ret, GDNF, and GFRa1 anti-sense riboprobes were generated as described (Srinivas et al., 1999b). For Wnt11 antisense riboprobe, the Wnt11 cdna clone pnk3 was linearized with XhoI and transcribed with T3 polymerase (Kispert et al., 1996). Results A Hoxb7/GDNF transgene can induce abnormal branching of ureteric bud-derived epithelia in vivo The Hoxb7 gene is expressed throughout the Wolffian duct and ureteric bud epithelium, and its promoter directs a similar pattern of transgene expression (Kress et al., 1990; Srinivas et al., 1999a,b). Using this approach, it was previously found that the expression of constitutively active forms of Ret could induce abnormalities in UB growth and branching (de Graaff et al., 2001; Srinivas et al., 1999b). Here, we sought to determine whether the ectopic expression of the Ret ligand GDNF would have similar effects. We first generated transgenic mice that carried a GDNF cdna under the direct control of the Hoxb7 promoter (Fig. 1A). Of 23 independently generated Hoxb7/GDNF transgenic founder mice examined at E18.5, two had multiple, collecting duct-derived renal cysts (Figs. 1B E) and one of these also contained branched, ureteric bud-derived epithelial tubules associated with the male reproductive tract, surrounded by undifferentiated mesenchyme (Figs. 1F, G), similar to abnormalities previously observed in mice expressing a constitutively active form of Ret (Srinivas et al., 1999b). As expected, the UB epithelium in these mice expressed high levels of GDNF mrna (data not shown). The lack of any obvious defects in the remaining 21 transgenic animals appeared to be due to lack of, or very low, expression of the transgene (data not shown). While the phenotype of the two affected mice supported the hypothesis that misexpression of GDNF could disrupt normal UB branching morphogenesis, the low frequency of expression of this transgene led us to use a different approach for further studies. A Doxycycline-inducible transgene system for regulated expression in the Wolffian duct and ureteric bud To examine in more detail the effects of GDNF misexpression at different stages of excretory system development, we employed the Tet-on system of inducible transgene expression (Furth et al., 1994; Gossen et al., 1995). We developed a transgenic line expressing the Doxycycline (Dox)-dependent transcriptional activator rtta2 S -M2 (Urlinger et al., 2000) under the control of either the Hoxb7 promoter (Hoxb7/rtTA) or the Axin2 promoter (Axin2/rtTA) (Fig. 2A). Axin2, like Hoxb7, is expressed in the Wolffian duct and UB epithelia but not in the surrounding mesenchyme (Jho et al., 2002 and data not shown). To test the efficacy of the Hoxb7/rtTA transgene, we first used a teto/lacz reporter transgene (Fig. 2C) (Furth et al., 1994). Adult mice carrying both transgenes, which were maintained on Dox during preand postnatal life, showed strong expression in the renal collecting duct system (Fig. 2D). Similarly, Dox-treated bi-transgenic embryos displayed strong expression of h- gal throughout the Wolffian duct, ureteric bud, vas deferens, epididymis and seminal vesicles (Figs. 3A, Fig. 4. Growth and branching of ectopic ureteric buds at E14.5 (A F) and E18.5 (G L). A, control Hoxb7/rtTAtetO/lacZ bi-transgenic male, showing expression of h-gal in Wolffian (asterisk) and Mullerian (arrow) ducts, as well as UB branches in the kidney (k). a, adrenal; b, bladder; s, sympathetic ganglia; t, testis. B, Hoxb7/rtTABiTetO/lacZ/GDNF bi-transgenic male with multiple ectopic UBs, most ending blindly (white arrows) but some connecting to the kidney (black arrow). Two ureters in the correct position are indicated by asterisks. C, dorsal view of the specimen in B, showing connection of the ureters to the medial pole of the kidneys, and connection of several ectopic UBs (black arrow) to the posterior pole. D, female Axin2/rtTABiTetO/lacZ/GDNF bitransgenic with multiple ectopic UBs, many of them branching. E, close-up of branching ectopic buds from D. g, gonad. F, section of specimen in E, showing branching epithelial buds along the Wolffian duct (asterisks), and expression of h-gal in a subset of cells in the Wolffian duct and ectopic buds, but not in the Mullerian duct (arrows). Note the absence of nephrogenesis in the mesenchyme surrounding the ectopic UBs. G, control Hoxb7/rtTAtetO/lacZ bi-transgenic, showing expression in the vas deferens (v), epididymis (inset), ureters (u) and kidneys. H, Axin2/rtTABiTetO/lacZ/GDNF bi-transgenic with many ectopic UBs (white arrows) that have undergone several rounds of branching, as well as multiple hydroureters (u). I, enlarged back view of H, showing extensively branched ectopic UBs on the epididymis. J, Hoxb7/rtTABiTetO/lacZ/GDNF bi-transgenic male with a triple ureter (red arrows) and a triplex kidney. K, section of the multiplex kidney in J, showing multiple ureters and division of the kidney into three segments, separated by nephrogenic zones (arrows). L, a female Hoxb7/rtTABiTetO/lacZ/GDNF bi-transgenic with ectopic h-gal-positive buds (arrows) attached to the uterus and oviduct, but no remaining Wolffian duct. LV, close-up of oviduct in L. LW close-up of a branched bud attached to uterus in L. Lj, section of LV, showing that the h-gal-positive branched buds are not continuous with the oviduct epithelium (arrows). Scale bars, 0.5 mm.

7 76 R. Shakya et al. / Developmental Biology 283 (2005) A, 4G). In contrast, no h-gal expression was observed in singly transgenic teto/lacz embryos treated with Dox, or in bi-transgenic embryos in the absence of Dox (data not shown). Thus, the Hoxb7/rtTA transgenic mice represent a useful tool for conditional expression of transgenes in cells of the Wolffian duct and ureteric bud lineages, during both pre- and postnatal development. Fig. 5. Tip-specific gene expression patterns are maintained in ectopic ureteric buds. A and B, in E14.5 Hoxb7/rtTABiTetO/lacZ/GDNF bitransgenic fetuses that also carry a Hoxb7/GFP transgene (Srinivas et al., 1999a), GFP is expressed throughout the ectopic UBs that formed along the Wolffian duct. Arrows point to some of the highly branched ectopic buds, and arrowheads in B point to an ectopic ureter. AV, whole mount in situ hybridization for Ret in the same specimen shown in A. Ret is expressed only at the tips of the ectopic buds (arrows). C, Ret in situ hybridization to a section of a similar bi-transgenic embryo, showing tip-specific expression in the ectopic buds (arrows) along the Wolffian duct (arrowheads). BV, whole mount in situ hybridization for Wnt11 in the same specimen shown in B. Arrows, specific expression at the ectopic UB tips; arrowhead, specific expression at tip of the ectopic ureter. D, Raldh2 is expressed in stroma (arrows) surrounding the UBs within the kidney (k), but not in the mesenchyme surrounding the ectopic UBs (asterisks); a, adrenal; g, gonad; k, kidney. Scale bars, 0.3 mm. Expression of GDNF in the Wolffian duct epithelium causes supernumerary ureteric buds to emerge and branch outside the kidney To examine the effects of GDNF misexpression on urogenital development, we crossed the Hoxb7/rtTA or the Axin2/rtTA activator strain with a BiTetO-lacZ-GDNF transgenic responder strain, which carries GDNF and lacz coding sequences under the control of a Dox-inducible, bi-directional promoter (Fig. 2B and Kholodilov et al., 2004), and administered Dox to the mothers beginning at E8.5. At E11.5, the Hoxb7/rtTABiTetO/lacZ/GDNF bitransgenic embryos displayed multiple ectopic buds (average 7, range 0 21, N = 24) along the entire length of the Wolffian duct (Figs. 3B, C). The variation in bud number correlated with variability in the level of GDNF transgene expression, as reflected by the intensity of h-gal staining (Fig. 3 and data not shown). The BiTetO/lacZ/GDNF transgene was expressed in most or all cells of the Wolffian duct and early UB, unlike the mosaic expression observed at later stages (see below). In bi-transgenic male embryos at E14.5, ectopic buds were observed along the vas deferens and epididymis, derivatives of the Wolffian duct. These buds had grown larger since E11.5, and many had undergone secondary branching (Figs. 4B F). Most of the ectopic buds ended blindly, but some connected to the metanephric kidney (e.g., Figs. 4B, C). By E18.5, branching of the ectopic buds was often even more extensive (Figs. 4H, I), similar to what was seen in Hoxb7/GDNF transgenics (Figs. 1F, G). In males, the buds remained connected to the Wolffian duct derivatives. In females, most of the Wolffian duct had degenerated (as normally occurs by E15.5 in the absence of male hormones, Kobayashi and Behringer, 2003) but the h-galpositive buds remained as isolated epithelial structures, embedded in the mesenchyme of the oviduct and uterus (Fig. 4L). Presumably, these buds survived while the duct degenerated, because they had differentiated from Wolffian duct into ureteric bud epithelium (a conclusion supported by marker gene expression patterns, as described below), and were thus refractory to the mechanism that causes degeneration of the Wolffian duct epithelium in females. In addition to the numerous ectopic buds, bi-transgenic fetuses usually had a correctly-positioned ureter that connected normally to the bladder and the medial pole of the kidney (e.g., Figs. 4B, C). In some cases (37%, N = 26), there were two or three ureters on one side (Figs. 4H, J, K), each associated with a separate renal pelvis, resulting in a duplex or triplex kidney (Figs. 4J, K). Frequently (54%, N = 26), one or more of the ureters was an expanded hydroureter (Figs. 4H, J, K), which in males could be seen to terminate on the sex duct instead of the bladder, thus preventing proper drainage. This observation is consistent with the model of Mackie and Stephens (1975), in which abnormal positioning of the ureteric bud along the anterior posterior axis of the Wolffian duct is thought to result in an

8 R. Shakya et al. / Developmental Biology 283 (2005) ectopic ureter that terminates abnormally either inside, or outside of the bladder. Unexpectedly, in many bi-transgenic females (typically those with the most ectopic UBs), the uterus was missing in whole or in part (data not shown). As the uterus is derived from the Mullerian duct (which normally develops parallel to the Wolffian duct Fig. 4A), this suggests that the Mullerian ducts either did not form fully, or degenerated. As Ret is not expressed in the Mullerian duct, it is unlikely that the GDNF secreted by the nearby Wolffian duct and ectopic ureteric buds could have acted directly on the Mullerian duct epithelium. An alternative possibility is that the ectopic ureteric buds secreted a factor(s) that inhibited Mullerian duct formation or persistence either directly, or indirectly, but this hypothesis remains to be further explored. Ectopic buds express ureteric bud markers with proper proximal distal patterning in the absence of metanephric mesenchyme or renal stroma The buds that formed at inappropriate locations were surrounded by mesenchymal cells, but this mesenchyme did not condense as it does around the normal UB tips, nor did it form epithelial vesicles (Figs. 1G, 4F, 4Lj). To determine if this mesenchyme had properties of metanephric mesenchyme or renal stroma, we examined the expression of the metanephric mesenchyme markers WT1 (Kreidberg et al., 1993) and GDNF and the stromal markers FoxD1 (Hatini et al., 1996) and Raldh2 (Niederreither et al., 2002). None of these markers were expressed by the mesenchymal cells surrounding the ectopic buds (Fig. 5D and data not shown). Thus, ectopic ureteric buds do not induce the surrounding Fig. 6. GDNF misexpression in developing collecting ducts of Hoxb7/rtTABiTetO/lacZ/GDNF bi-transgenic mice, and its effects of on kidney histology at E18.5. A, Wild type kidney section. B, enlargement of wild type papilla. C, enlargement of wild type cortex and nephrogenic zone. D, Section of a bitransgenic kidney, showing normal overall organization but a few scattered cysts (*). E, enlargement of papilla from D, showing several cystic, branched collecting ducts. F, enlargement of cortex and nephrogenic zone, showing a cortical cyst (*). G, section of a different bi-transgenic kidney, showing hydroureter (U), lack of a normal-shaped pelvis (p), disorganization of the medullary region and numerous cysts (*). H, enlargement of medulla from G, showing several abnormally branched and cystic collecting ducts (*). I, enlargement of cortex, showing several cystic collecting ducts (*). J, expression of GDNF mrna detected by in situ hybridization in a bi-transgenic kidney. K, enlargement of the medulla of J, showing strong expression of GDNF in a subset of collecting duct cells. L, enlargement of the cortex of J, showing strong expression of GDNF in a subset of collecting duct cells, and weaker expression of endogenous GDNF in the peripheral mesenchyme (bracket). A I, H and E stain. Scale bars, 0.3 mm. A, D, G at same magnification; B, C, E, F, H, I, K, L at same magnification.

9 78 R. Shakya et al. / Developmental Biology 283 (2005) mesenchymal cells to form normal metanephric mesenchyme or renal stroma. This is consistent with the fact that in normal development, metanephric mesenchyme appears to be specified before the invasion of the ureteric bud (Brophy et al., 2001; Grieshammer et al., 2004; Kreidberg et al., 1993). To examine patterning of the ectopic buds, we analyzed expression of three genes, Ret, Wnt11, and GFRa-1, which are expressed in tips of the normal branching UB epithelium, but not in the UB trunks or the Wolffian duct at E14.5. The expression of all three genes was restricted to the tips of the ectopic buds (Fig. 5 and data not shown). Thus, the ectopic buds resemble normal ureteric buds not only in their ability to branch but also in their pattern of tip-specific gene expression. The expression of UB markers with a normal proximal distal pattern appears to be intrinsic to the UB branching morphogenesis program and does not require inductive signals from the metanephric mesenchyme or stroma. This conclusion was further supported by studies in which wild type E11.5 ureteric buds were cultured in a system that allows branching morphogenesis in the absence of mesenchymal cells (Qiao et al., 1999). The expression of Ret (Qiao et al., 1999) and Wnt11 (B. Lu and F.C., unpublished data) was also restricted to the tips of the branched ureteric bud-derived epithelia in these cultures. The kidneys of the bi-transgenic mice displayed a range of phenotypes: many appeared normal, while a few displayed defects similar to, although less severe than, those observed in the Hoxb7/GDNF mice shown in Fig. 1. Aside from multiple ureters and kidneys, described above, 30% had cortical collecting duct cysts (Figs. 6F, I) and/or dilated medullary collecting ducts (Figs. 6E, H), 9% had abnormal branching of medullary collecting ducts (Fig. 6H), and 23% had a misshapen and dilated pelvis (Fig. 6G). These results are consistent with our preliminary observations (Fig. 1) that misexpression of GDNF in the UB epithelium can induce abnormalities in its growth and branching. The heterogeneity in phenotype, and the generally mild defects, seemed to be a consequence of the variable and mosaic expression of the BiTetO/lacZ/GDNF transgene in the kidney, as revealed by in situ hybridization for GDNF mrna (Figs. 6J L) or staining for h-gal (e.g., Fig. 4F). This mosaic expression was a specific property of the BiTetO/lacZ/GDNF responder transgenic line, and apparently a consequence of transgene methylation (data not shown), a well-established phenomenon (Allen et al., 1990; Engler et al., 1991). In contrast, the Hoxb7/rtTA activator was uniformly expressed in the renal collecting ducts, as shown by control experiments with the teto/lacz responder transgene (Fig. 2D). Since the GDNF transgene was expressed in a small proportion of UB cells within the kidney, and endogenous GDNF was expressed normally in the peripheral mesenchyme of these kidneys (Fig. 6L, bracket), we suspected that the overall distribution of GDNF might be only minimally altered. Thus, according to a model in which gradients of secreted GDNF pattern the growth and branching of the UB, the collecting duct system might be relatively normal in these mice because, at the periphery of the kidney, GDNF was still produced mostly in the mesenchyme and only in a few UB cells. To further test this model, we next generated animals that expressed GDNF solely in the ureteric bud epithelium and lacked any mesenchymal GDNF expression. GDNF expressed by the ureteric bud epithelium supports kidney development in the absence of mesenchyme-derived GDNF To produce mice lacking endogenous GDNF in the metanephric mesenchyme and expressing it only in the Renal collecting duct abnormalities induced by misexpression of GDNF Fig. 7. Rescue of kidney development in gdnf / mice by transgenic misexpression of GDNF in the ureteric bud. A, lack of ureters or kidneys in a newborn gdnf / mouse (a gdnf lacz knock-in null allele, Sanchez et al., 1996), stained for h-gal. g, a segment of gut; a, adrenals; t, testes; b, bladder. B, a control gdnf lacz heterozygote, with normal kidneys expressing h-gal in the pattern of gdnf. k, kidney; ur, ureter; ut, uterus; o, ovary. C, a newborn rescued gdnf / mouse carrying the Hoxb7/rtTA and BiTetO/ lacz/gdnf transgenes. On the left is a duplex kidney, and on the right is a single kidney. h-gal stain in the ureters and ectopic UBs (arrows) is due to expression of the BiTetO/lacZ/GDNF transgene, while h-gal in the kidney is from both the gdnf lacz allele and the BiTetO/lacZ/GDNF transgene. D, a different rescued newborn gdnf / mouse, with two kidneys and ureters (not stained for h-gal). Arrows, ectopic UBs. Scale bar, 0.3 mm.

10 R. Shakya et al. / Developmental Biology 283 (2005) Wolffian duct and ureteric bud lineages, we introduced a gdnf null mutation. Hoxb7/rtTA, gdnf +/ mice were crossed with BiTetO/lacZ/GDNF, gdnf +/ mice and the resulting offspring were examined at term. The gdnf / offspring that did not inherit both transgenes either had no kidneys (60%) (Fig. 7A) or tiny kidney rudiments smaller than the adrenal gland (40%) (Fig. 8A), as previously observed (Moore et al., 1996; Pichel et al., Fig. 8. Histology of rescued kidneys and expression of GDNF, Ret and FoxD1. A, a rudimentary kidney from a gdnf / mouse. B, the rescued kidney shown in Fig. 7D. Note the similar organization to wild type (Fig. 6A), with well-shaped pelvis (p), medulla, cortex and nephrogenic zone. a, adrenal. C, enlargement of medulla from B, showing a few dilated collecting ducts (*). D and E, enlargements of two regions of the nephrogenic zone of the rescued kidney in B. The arrowhead in D indicates the thickened peripheral stromal layer, and inset DVshows that these cells express the stromal marker FoxD1. The green arrow in D points to a glomerulus, and the double arrow in E to a bifurcated UB at the periphery. F, a different rescued kidney, with normal overall organization comparable to wild type (Fig. 6A), but smaller in size. G, enlargement of papilla and pelvis from F, showing several dilated collecting ducts (*). H, enlargement of cortex and nephrogenic zone from F. The asterisks indicate UB tips, the green arrow a glomerulus, and the black arrowhead a thickened stromal layer. I J, in situ hybridization for GDNF in the rescued kidney shown in B. I, transgene-encoded GDNF is strongly expressed in a subset of collecting duct cells in the medulla (black dots). IV, enlargement showing several ducts with GDNF-positive cells. J, in the cortex, transgene-encoded GDNF is expressed in a smaller subset of collecting duct cells than in the medulla (small arrows), while endogenous GDNF, usually expressed in the peripheral mesenchyme (bracket), is absent (compare to Fig. 6L). JVand JW, enlargements showing GDNF-expressing cells in collecting ducts (small arrows). The double arrow in JVindicates a T- shaped UB. K, in the same kidney, Ret is expressed primarily in the UB tips at the periphery of the kidney, despite the absence of GDNF expression in the mesenchyme. L, enlargement of K. Scale bars in IV, JV, JW are 0.05 mm; all others are 0.3 mm. D, DV, E at same magnification as C.

11 80 R. Shakya et al. / Developmental Biology 283 (2005) ; Sanchez et al., 1996). In contrast, of the eight gdnf / offspring that inherited and expressed both transgenes, three had kidneys that were essentially normal in shape and position (Figs. 7C, D), although they were somewhat smaller than those in wild type or gdnf +/ mice (Fig. 7B), and two of these had a duplex kidney and ureter on one side. Thus, the transgenes had rescued kidney development to a substantial degree in these gdnf / mice. Of the remaining 5/8 mice, two more had kidneys that were significantly larger than those in gdnf / mice, but were misplaced or obviously cystic (not shown), and three had no kidneys. All of the gdnf / mice, including those with rescued kidney development, still displayed aganglionosis of the gut (data not shown), as was expected since the Hoxb7/rtTA transgene is not expressed in the gut; this provided a phenotypic confirmation of the gdnf / genotype. The rescued kidneys were remarkably close to normal in histoarchitecture, with only minor histological abnormalities. They contained a normally shaped pelvis, a distinct medulla, a cortex with abundant glomeruli, a peripheral nephrogenic zone containing comma and S- shaped bodies, and an outer layer of FoxD1-expressing stromal cells (Figs. 8D, DV, E, H). The stromal cell layer was thicker than normal, as has been seen in some mutants with defects in ureteric bud growth (Batourina et al., 2001; de Graaff et al., 2001). Most importantly, the developing collecting duct system showed many of the hallmarks of normal patterning, with the ureteric bud tips growing outward at the periphery of the kidney, and clearly displaying the terminal bifurcation typical of UB branching (Figs. 8E, J). The only obvious abnormalities in the collecting system were seen in the medulla of two rescued kidneys, where a few collecting ducts were dilated (Figs. 8C, G). In addition, there were numerous ectopic ureteric buds along the reproductive tracts, as expected (Figs. 7C, D). In situ hybridization confirmed that endogenous, mesenchymal GDNF was absent (Fig. 8J compare to Fig. 6L), and that the GDNF transgene was expressed specifically in collecting duct cells (Figs. 8I, IV, J, JV, JW). The transgenic GDNF mrna was expressed in only a small proportion of medullary collecting duct cells (Figs. 8I, IV) and in even fewer cortical collecting duct cells (Figs. 8J, JV, JW). However, the level of GDNF mrna in these cells appeared higher than it is in normal mesenchyme cells. Thus, it appears that the GDNF secreted by a small number of UB cells in these mice was sufficient to support renal development. Despite the absence of endogenous GDNF in the nephrogenic zone, the strong expression of Ret and GFRa-1 mrnas at the UB tips was maintained (Figs. 8K, L and data not shown). Thus, as in the ectopic ureteric buds induced to form outside the kidney, the expression of UB tip-specific markers in the kidney does not require a localized, mesenchymal source of GDNF. Discussion Mesenchymal epithelial interactions are an important mechanism for patterning the growth of organs that form by branching morphogenesis (Hogan and Yingling, 1998; Vainio and Lin, 2002). In the developing excretory system, GDNF is one of the factors secreted by mesenchymal cells and believed to play a key role in patterning epithelial growth and branching. Genetic studies in mice, in combination with organ culture studies using exogenous GDNF, have led to a model in which the localized, mesenchymal expression of GDNF is critical not only for positioning the site of outgrowth of the UB from the Wolffian duct, but also for patterning the branching and growth of the UB within the developing kidney (Lechner and Dressler, 1997; Sariola and Saarma, 2003; Sariola and Sainio, 1997; Vainio and Lin, 2002), a process that is important for normal renal histoarchitecture (Al-Awqati and Goldberg, 1998; Oliver, 1968). Here, we have tested the requirement for GDNF as a paracrine, chemoattractive factor, by altering its site of expression during excretory system development. Our findings confirm the conclusion that the localized expression of GDNF along the posterior Wolffian duct is important for the formation of a single, correctly placed ureter. They also support the role of GDNF as a ramogen, that is, a factor that promotes the branching of an epithelial tube. However, they do not support the view that the localized expression of GDNF is critical for patterning the growth of the ureteric bud within the kidney, since the replacement of mesenchymal GDNF with UB-derived GDNF had only minor effects on kidney development. In normal embryogenesis, a single ureteric bud evaginates from the Wolffian duct and grows dorsally into the metanephric mesenchyme, which expresses GDNF. The misexpression of GDNF throughout the Wolffian duct epithelium led to the outgrowth of as many as 20 ectopic buds all along the Wolffian duct, demonstrating that the entire Wolffian duct is competent to form ureteric buds. Our findings are consistent with previous in vitro studies showing that GDNF-soaked beads placed near the posterior Wolffian duct can elicit the formation of supernumerary buds (Brophy et al., 2001; Sainio et al., 1997; Towers et al., 1998), and they extend previous in vivo observations that mutations that slightly expand the domain of GDNF expression in the nephrogenic cord can alter the sites of UB outgrowth (Grieshammer et al., 2004; Kume et al., 2000). Thus, the normally restricted expression site of GDNF plays an important role in determining the correct position of ureteric bud formation (Lechner and Dressler, 1997; Sainio et al., 1997). However, it is important to note that, although most embryos lacking GDNF, Ret or Gfra1 fail to make a ureteric bud, some of them do form a UB in approximately the correct position, which grows towards the metanephric mesenchyme, although its elongation and branching are severely impaired (Moore et al., 1996; Pichel et al., 1996; Schuchardt et al., 1996). We also observed that,

12 R. Shakya et al. / Developmental Biology 283 (2005) even in transgenic fetuses with multiple ectopic buds along the Wolffian duct, only the bud(s) that formed at or near the correct position went on to form a ureter with a normal connection to the kidney. Therefore, even when GDNF is not correctly localized, other signals limited to the posterior region, and probably produced by the metanephric mesenchyme, appear to contribute to the positioning and outgrowth of the ureter. One important difference from previous studies is that, in our experiments, GDNF was produced by the Wolffian duct itself, that is, by the same epithelial tube that expresses the GDNF receptors Ret and Gfra1. When the source of GDNF is a bead or a population of mesenchymal cells outside the Wolffian duct, it might generate a gradient of GNDF, which could guide the ureteric bud towards the source of GDNF. However, GDNF secreted by the Wolffian duct is unlikely to form the same gradient it would more likely form an inverted gradient, with the highest levels of GDNF close to the duct. Thus, the growth of buds away from the Wolffian duct does not require a localized external source of GDNF, but merely that the Wolffian duct cells be exposed to GDNF. It is also intriguing that, despite the uniform expression of the BiTetO/lacZ/GDNF transgene along the Wolffian duct epithelium (Fig. 3C), the buds formed as discrete tubular outgrowths, rather than as a continuous swelling of the entire Wolffian duct. This suggests that a region of the Wolffian duct epithelium that begins to bud in response to GDNF sends out inhibitory signals that repress budding in the adjacent region of the duct. Unlike the ectopic buds induced by GDNF beads in vitro, which do not branch again after their initial evagination from the Wolffian duct (Brophy et al., 2001; Sainio et al., 1997), many ectopic buds in the transgenic mice continued to grow and branch repeatedly. This is likely due to the continued expression of GDNF by the epithelium of the ectopic buds. Many of these buds were quite far from the developing kidney (and therefore not exposed to factors secreted by the metanephric mesenchyme), and the mesenchymal cells surrounding them did not display features of metanephric mesenchyme or renal stroma. Therefore, it appears that the GDNF expressed by the buds themselves is inducing their continued branching. In contrast, when FGF-7 (which also signals through a tyrosine kinase receptor) was expressed in the Wolffian duct epithelium, it induced uniform swelling of the duct but not budding or branching (unpublished data). This supports the view that GDNF has the specific effect of inducing the Wolffian duct and UB epithelium to branch (Davies et al., 1999; Pepicelli et al., 1997; Towers et al., 1998; Vega et al., 1996). The branching of the ectopic UBs did not follow the stereotypical pattern that occurs within a normal developing kidney. Atypical patterns of branching morphogenesis is also seen in other situations in which the UB is induced to branch outside of its normal relationship with metanephric mesenchyme, for example, when cultured in an artificial matrix (Qiao et al., 1999) or recombined with lung mesenchyme lung (Kispert et al., 1996; Lin et al., 2003). Therefore, the metanephric mesenchyme imposes pattern on the growth and branching of the ureteric bud. Whether the localized expression of GDNF within the metanephric mesenchyme is a critical determinant of this pattern is a question we address below. One aspect of ureteric bud patterning that we found to be independent of interaction with the metanephric mesenchyme was proximal distal differentiation into tip vs. trunk. It was previously observed that tip-specific expression of Ret and Wnt11 was maintained in E11.5 UBs recombined with lung mesenchyme (Kispert et al., 1996), although it was possible that localized expression of GDNF in the lung mesenchyme accounted for this pattern, as both Ret and Wnt11 are GDNF target genes (Pepicelli et al., 1997). In our experiments, the normal tip-specific expression of Ret, Gfra1 and Wnt11 in ectopic buds from the Wolffian duct, as well as in isolated wild type UBs cultured without mesenchyme, showed that this pattern is an intrinsic property of the developing UB. It does not depend on a localized source of GDNF, as GDNF was expressed throughout the epithelium of the ectopic buds, or throughout the medium surrounding the cultured UBs. Nor do ureteric bud tip and trunk cells represent two entirely separate cell lineages with different patterns of gene expression, as we have found that some tip cells give rise to trunk cells during normal UB growth (Shakya et al., 2005). How ureteric bud cells know whether they are in the trunk or tip is an interesting problem that remains to be solved. To investigate the role of GDNF in patterning ureteric bud branching morphogenesis within the developing kidney, we first examined kidneys from newborn transgenic mice on a wild type background, which misexpressed GDNF in the UB epithelium while still expressing endogenous GDNF in the metanephric mesenchyme. The extent of renal defects varied considerably, from a few severely affected cases with multiple branched, UB-derived cysts (e.g., Fig. 1), to others with fewer and smaller cysts (Fig. 6), to many apparently unaffected cases. This wide range of severity was apparently a consequence of the variable levels of GDNF transgene expression. The defects observed in the more severely affected cases were similar to those previously observed in transgenic mice that expressed ligand-independent forms of Ret throughout the UB (de Graaff et al., 2001). This confirms that the misexpression of GDNF can indeed perturb the pattern of UB branching morphogenesis, although this may be due to the elevated overall level of secreted GDNF rather than to the altered spatial pattern of expression. While these observations were consistent with ability of GDNF to stimulate epithelial branching and growth, they did not provide a clear answer regarding its importance as a chemoattractive factor. The relatively normal patterning of the collecting system in most of the transgenic kidneys might imply that the site of GDNF expression is not important for UB patterning, or it might